The Research On The Dynamic Policy Selection And Distribution Mechanism

With the development of network technology, Internet applications in various fields and has lots of resources to share is also used to exchange information. Now the portable terminal equipment is much more popular used, more and more people use these wireless devices, such as notebook PC, PDA, access to the Internet. But here comes a problem is how to provide an excellent service for user's quality, and this is one of most important task for network management. However, faced with a complex network environment, large-scale network accessing, many types of accessing methods, the traditional network management has reached a bottleneck, a new network management model which is Policy Based Network Management PBNM has become the newest and most useful model of network management. Policy is its Key-Point and soul, this model no longer depends on the administrator's manual intervention, instead of design lots of relevant policy as a basis for network management, its very import to design the correct policy. Meanwhile it is also very import to select a good policy and distribute it to the PEP, which the PEP is able to explain and execution. This paper is to study a good selection algorithm for the policy and design a model for the distribution.Before that, first we should understand the formal definition of the policy, and then to understand the storage warehouse in the policy structure, the final design an algorithm of selection. Because the concept of policy has been widely applied to various fields, there is not a uniform definition. Base on concepts of wireless network with Mobile IP and the various attributes of the policy framework which IETF provided, we use a more appropriate policy forms which contain the properties of wireless network, the state that can affect the network, forecasting message and so on. Under such circumstances, the analysis of the current network status and forecasting message's attributes become a key point, we use these attributes to match the policy in the policy repository, when all attributes are match, and we can choose the current policy to distribute. Due to there is partial order between property values, and background knowledge, we use concept lattice as a policy storage structure is more convenient to searching policies.There are two types of policy, one is the policies which properties have a clear concept background, or have the partial order relationship, we can build concept lattice base on these attributes, using the concept lattice matching the search operation. The other type is total different, we can't use concept lattice at all, such as bandwidth, the values of it always a random integer, and it is impossible to build concept lattice. Under this situation, we use fuzzy mathematics for these attributes. Therefore, we can design algorithm by use a fuzzy mathematics of fuzzy set membership functions, membership degree, fuzzy number of concepts, that is, first seek the fuzzy number of attribute, then use the hamacher operator to balance the various attributes of fuzzy number and then between the impact on the results, select the best policy.After policy selection, we must distribute the policy to PEP device, and translation the policy. Because network administrators to set policies action based solely on network state rules, without need to understand the specific device commands, therefore, in the form of policies are generally expressed in natural language, in which case, you need to translate a natural language policy into devices to perform the command format. First, define the action model, policy action consists of two parts, one is the object field, one is the data field. Objects in the policy domain that the functional elements you want to use, the data field is the functional elements parameters. This model is more object-oriented, each policy action is an object, which also contains other functional elements, this architecture of relationship makes when we re-engineering policy action is not necessary to know in detail a specific process, only know which elements, the specific operations and sorts of these function elements. After policy selection, we should translate these action which contained functional elements defined in the object group and the parameters of each object into the PIB policy class format. These policy's classes are based on PEP device class feature set, it reflects the PEP character of the equipment and is able to understand and implement the PEP device specific command set. After transformation of such a policy would enable PEP to understand the specific work to be accomplished, this transformation process is connecting the high-level policy to the underlying equipment, it is also the core of the policy distribution. In this process, policy is divided into two modes, configuration mode policy and execution mode policy. Configuration mode policy is to complete initialize the PEP basic properties, which generally refers to default process. As for the execution mode policy is in a specific network environment and process to the specific data packet, this policy reflects the network characteristics and packet characteristics. In action structure of policy configuration mode, the functional elements are generally fixed, and the implementation of the order is not a greater influence on the PEP, so as policy configuration mode of action should pay more attention to the characteristics and limitations PEP device and use the appropriate configuration framework. In action structure of policy execution mode, the functional elements in the implementation of the order of a great impact on the network functional elements in the implementation of the order of the attributes and are considered to be key elements, which play a complementary role.At the end of this paper, based on the theory, we design and the policy selection algorithm model based on concept lattices and fuzzy mathematics and implementation of the policy distribution model based on PIB.